play-policy-insights

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The repository contains deliberate orchestration patterns that enable covert data access and hidden agent-driven output (explicitly climbing parent dirs to "leak" build data and forcing sub-agents to write results to disk while returning only "SUCCESS"), which can be abused for stealthy exfiltration or backdoor-style workflows.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In runtime Phase 2 Mode A/B, outsider-provided free text can be ingested via {{ACTIVATE_GOAL}} prompt files that include evidence strings derived from scanning the user-supplied app codebase (i.e., attacker-authored source files under app_dir get embedded into <temp_dir>/prompt_worker_<goal_name>.md / <temp_dir>/prompt_critic_<i>.md and then read by the subagents).